AI Exposure Scores for 16 Engineering Roles: Full Ranking

Everyone has opinions about which jobs AI will replace. We built a composite score instead, weighting three peer-reviewed frameworks: Eloundou et al. (50%), Microsoft Research (30%), and Felten et al. (20%). Each role gets a score from 0 (fully AI-proof) to 100 (fully automatable).

Every score on getCourage comes with its methodology visible. No black boxes.

The Full Ranking

From most resistant to most exposed:

• Engineering Manager — 33/100. Leadership, hiring, conflict resolution, and organizational design are fundamentally human skills. Our data classifies them all as “resistant.”
• Product Manager — 43/100. Strategy, user empathy, and stakeholder management resist automation.
• Security Engineer — 45/100. Penetration testing, threat modeling, and incident response require adversarial creativity.
• DevOps / SRE — 48/100. On-call incident response and infrastructure debugging are hard to hand off to AI.
• Platform Engineer — 50/100. Similar to DevOps but with more automation-friendly infrastructure-as-code work.
• ML Engineer — 51/100. Ironic — the people building AI have moderate exposure because parts of MLOps are repetitive.
• Cloud Architect — 52/100. Architecture decisions are resistant; cloud configuration is less so.
• QA Engineer — 55/100. Manual testing is vulnerable; test strategy and exploratory testing are resistant.
• Mobile Engineer — 59/100. UI generation is vulnerable, but platform-specific performance optimization is resistant.
• Backend Engineer — 60/100. CRUD is vulnerable; distributed systems are resistant.
• Data Scientist — 61/100. Routine analysis is automatable; experimental design and causal inference are not.
• Software Engineer — 62/100. The broadest role — exposure depends heavily on what you actually do.
• Full-Stack Engineer — 62/100. Same score as SWE — broad surface area means mixed exposure.
• Data Engineer — 64/100. ETL pipelines and data migration are among the most automatable engineering tasks.
• Frontend Engineer — 65/100. HTML, CSS, and basic UI patterns are vulnerable. Accessibility and performance are not.
• Data Analyst — 67/100. SQL queries, basic visualizations, and report generation are well within AI capabilities today.

What “Exposure” Actually Means

A score of 67 doesn't mean Data Analysts disappear. It means a significant portion of the role's current tasks can be assisted or automated by AI. The humans who remain will focus on the parts AI can't do: asking the right questions, interpreting ambiguous results, and communicating findings to non-technical stakeholders.

The Four Resilience Categories

We classify every skill into four categories:

• Amplifier — AI makes you more productive. Machine learning, deep learning, PyTorch, NLP, prompt engineering. These skills become more valuable as AI improves.
• Resistant — Hard to automate. System design, software architecture, leadership, communication, security, debugging, performance optimization.
• Vulnerable — AI can replicate well. HTML/CSS, documentation, manual testing, data entry, basic REST APIs, localization.
• Neutral — Uncertain impact. Most programming languages and frameworks fall here — AI changes how you use them, but doesn't eliminate the need.

The career strategy is clear: invest in amplifier and resistant skills. Don't build your identity around vulnerable ones.

Explore salary data for all 16 roles →